The world we live in is a connected one where consumers interact with personal and business services all day long. The advent of virtual assistants like Siri, Cortana, Alexa and Google Assistant have created ubiquitous access, even when a screen or device is not at hand.
In company and brand interactions, customers today expect to be able to speak, or type, their issues in natural language and get their problems solved quickly. This is placing increased demands on natural language and other human-computer interaction technologies to understand what you say or type, and tell you what to know or do.
But in spite of the popularity and prevalent use of natural language in customer experience, our research shows that NL technology on its own has a 40 percent failure rate in identifying actionable intent. This makes NL on its own an incomplete solution for today’s complex and demanding consumer.
Customer journeys take place across devices and channels, and customers expect the steps along these journeys to be connected, seamless, and productive. Even if a system could understand language requests perfectly, no customer wants to repeat steps or start over when they enter a new channel.
For example, consider a scenario whereby a customer is attempting to pay for a purchase when her credit card is declined. Thinking she may be over her spending limit, she logs into her online banking account where she learns that she is under her limit. Confused, she calls the 800 number and reaches a natural language speech system (IVR) that asks her, "How may I help you today?" She responds, "My card is not working."
Based on the words she used, a natural language system trained on thousands of similar utterances from other callers could return a set of probable intents:
- The customer reached her credit limit.
- The card is blocked due to suspected fraud.
- The card is damaged.
- The card has expired, and a new card has not been received.
- The card has been suspended due to a late payment.
The system would then disambiguate, or determine which of the intents is the correct one, by asking the customer more questions (somewhat annoying), presenting a long list of options (very annoying) or transferring to a human agent (time consuming and expensive).
This is where language understanding alone is not sufficient. Wouldn’t it be better if the credit card provider knew which intent was the right one based on the context of the customer’s actions? Since the bank can recognize the customer, it should know that a recent transaction was declined, and look into the reason for the decline. The bank could then respond appropriately, for example by saying “I see that you have a payment past due. Would you like to make a payment now?”
Prediction adds an extra artificial intelligence layer on top of natural language. A predictive platform that tracks and shares all customer activity across channels in real time would be able to deliver faster resolutions that result in superior experiences for customers. The solution can anticipate customer requests and follow the principle to “never ask the user something you should already now”. The system gains this predictive intelligence through real-time analysis of all customer activity, including web browsing behavior, mobile app usage, chat conversations, voice calls, shopping cart status, purchase history and customer profile data.
Predictive intelligence enables the higher-level conversation skill of active guidance. By continuously monitoring all contextual signals (in addition to listening to language signals), the system “realizes” what the user needs without waiting for the user to express it. The system does not wait for you to find something and, instead, offers you relevant information when you need it and guides you to the next best action.
Prediction extends beyond automated experiences. Customers who move into a chat session can now be greeted by an agent who is informed on who the customer is and what they want to do. Live chat agents and voice agents can be provided with recommended resolution paths for the customer, before the customer interaction even begins.
Natural language processing by itself solves only part of the problem of understanding intent and resolving customer needs. Predictive technologies reduce the need for someone to repeat, explain or clarify requests, thereby making spoken and typed language even more efficient as a means for getting things done.
Register or login for access to this item and much more
All Information Management content is archived after seven days.
Community members receive:
- All recent and archived articles
- Conference offers and updates
- A full menu of enewsletter options
- Web seminars, white papers, ebooks
Already have an account? Log In
Don't have an account? Register for Free Unlimited Access